Sequential Monte Carlo With Model Tempering

TL;DR

This paper introduces a sequential Monte Carlo method with model tempering that significantly speeds up Bayesian inference in complex macroeconomic models by reweighting and mutating posterior samples from simpler approximating models.

Contribution

It presents a novel SMC algorithm that efficiently accelerates Bayesian estimation for time series models with costly likelihood evaluations.

Findings

Runtime reductions of 27% to 88% achieved

Applicable to VAR with stochastic volatility and DSGE models

Demonstrates substantial computational efficiency improvements

Abstract

Modern macroeconometrics often relies on time series models for which it is time-consuming to evaluate the likelihood function. We demonstrate how Bayesian computations for such models can be drastically accelerated by reweighting and mutating posterior draws from an approximating model that allows for fast likelihood evaluations, into posterior draws from the model of interest, using a sequential Monte Carlo (SMC) algorithm. We apply the technique to the estimation of a vector autoregression with stochastic volatility and a nonlinear dynamic stochastic general equilibrium model. The runtime reductions we obtain range from 27% to 88%.